Boaz Barak | Polylogues

Channel:
United States Simons Institute for the Theory of Computing
Subscribers:
68,900
Published on ● Video Link: https://www.youtube.com/watch?v=rM1PeAWDNWQ


Duration: 0:00
1,156 views
16

In this episode of our Polylogues web series, Spring 2024 Science Communicator in Residence Ben Brubaker sits down with cryptographer and machine learning researcher Boaz Barak (Harvard and OpenAI). The two discuss Boaz’s path to theoretical computer science, the relationship between theory and practice in cryptography and AI, and specific topics in AI, including generality, alignment, worst- vs. average-case performance, and watermarking.



Other Videos By Simons Institute for the Theory of Computing


2025-02-18Knowledge is structured and domain-specific: lessons from developmental cognitive science
2025-02-18Virtual Lab of AI Scientists
2025-02-18You Know It Or You Don’t: Compositionality and Phase Transitions in LMs
2025-02-18Interpreting LLMs to Interpret the Brain
2025-02-18Language and thought in brains: Implications for AI
2025-02-18Dissociating language and thought in large language models
2025-02-18The Cognitive Boundaries of Language Models: Hallucinations and Understanding
2025-02-17How Do Transformers Learn Variable Binding?
2025-02-06Automating scientific discovery and hypothesis generation with language model agents
2025-01-26Resilience in Action
2025-01-24Boaz Barak | Polylogues
2024-12-24Debate: Sparks versus embers
2024-12-24Off-the-shelf Algorithmic Stability
2024-12-24Learning Theory of Transformers: Generalization and Optimization of In-Context Learning
2024-12-24First-Person Fairness in Chatbots
2024-12-24Panel on the future of scientific research and education
2024-12-24Generalization in the representations and computations of frontier language models.
2024-12-24Temporal Context in Brains and AI
2024-12-24Frame-shifting and Conceptual Blending: What do large language models have to say?
2024-12-24How machine learning is influencing protein engineering
2024-12-24Strong generalization from small brains and no training data